A CLASSIFICATION OF EQUIVARIANT GERBE CONNECTIONS BYUNGDO PARK AND CORBETT REDDEN Abstract. Let G be a compact Lie group acting on a smooth manifold M. In this paper, we consider Meinrenken's G-equivariant bundle gerbe connections on M as objects in a 2-groupoid. We prove this 2-category is equivalent to the 2-groupoid of gerbe connections on the differential quotient stack associated to M, and isomorphism classes of G-equivariant gerbe connections are classified by degree three differential equivariant cohomology. Finally, we consider the existence and uniqueness of conjugation-equivariant gerbe connections on compact semisimple Lie groups. Contents 1. Introduction1 2. Background and conventions3 2.1. Sheaves of 1-groupoids on the site of manifolds3 2.2. Relevant examples5 2.3. Cohomology and differential cohomology8 3. The 2-groupoid of equivariant bundle gerbe connections9 3.1. The definition 10 3.2. Basic properties 12 3.3. Fiber product morphisms 13 3.4. Trivializations and isomorphism classes 16 3.5. The induced simplicial gerbe 19 4. Gerbes on the quotient stack 19 4.1. The functor between the two models 21 4.2. Proof of equivalence 25 5. Applications 28 5.1. Equivariant holonomy 28 5.2. Equivariant gerbes over compact semisimple Lie groups 29 5.3. Higher gerbes 30 References 31 1. Introduction While there exist several models for S1-banded gerbes with connection, working directly with any of them necessarily involves technical details and subtleties, and the complexity grows rapidly should one consider p-gerbes for p > 1. However, there is a relatively simple classification of gerbe connections up to isomorphism, which is given by degree 3 differential cohomology (Deligne cohomology) and fits into the pattern 1 1 1 2 3 ∼ ∼ 1 ∼ (1.1) Hb (M) = C (M; S ); Hb (M) = BunS ;r(M)=∼= ; Hb (M) = Grbr(M)=∼= : 2 BYUNGDO PARK AND CORBETT REDDEN Because the groups Hb n(M) fit into short exact sequences involving differential forms and ordinary cohomology, one can use elementary techniques to get a strong foothold in the gerbe world. Assume now that G is a compact Lie group acting smoothly on a finite-dimensional manifold M. This paper's aim is to further develop the theory of G-equivariant gerbe 1 connections in a manner analogous to G-BunS1;r(M), the groupoid of G-equivariant S - bundles on M with G-invariant connection. We do so by considering two specific models. The first is denoted G-Grbr(M), the 2-groupoid of G-equivariant bundle gerbes on M with G-equivariant connection. Its objects, which were defined by Meinrenken [Mei] and further investigated by Sti´enon[Sti] and Tu{Xu [TX], are differential-geometric in the same way as Murray's bundle gerbes with connection [Mur]. We define a 2-categorical structure on G-Grbr(M) by adapting Waldorf's 2-groupoid structure on Grbr(M) in a straightforward way [Wal1]. The second model uses the language of higher stacks, or sheaves of 1-groupoids on the site of manifolds. For M 2 G-Man we consider the differential quotient stack ErG ×G M, which associates to every test manifold X the groupoid of principal G-bundles with con- nection (P; Θ) 2 BunG;r(X), together with G-equivariant map f : P ! M. In [Red1], the differential equivariant cohomology groups were defined as n n HbG(M) := Hb (ErG ×G M); suitably interpreted, and it was shown that they fit into short exact sequences (2.23){(2.25) ∗ ∗ involving the Borel equivariant cohomology groups HG(M; −) := H (EG×G M; −) and the ∗ equivariant differential forms ΩG(M) of Cartan{Weil. Furthermore, there are isomorphisms generalizing (1.1) in degrees one and two: 1 1 1 G 2 ∼ ∼ 1 HbG(M) = C (M; S ) ; HbG(M) = G-BunS ;r(M)=∼= : It was further shown in [Red2] that there is a natural equivalence of groupoids ∼= G-BunK;r(M) BunK;r(ErG ×G M) 3 for any Lie groups G; K. When combined with the properties of HbG(M), this strongly suggests that Grbr(ErG×G M) should be a valid model for G-equivariant gerbe connections. That hope is confirmed by the two primary results of this paper, Theorems 4.15 and 4.16. The first theorem states ∼= 3 (1.2) G-Grbr(M)=∼= −! HbG(M); that isomorphism classes of G-equivariant bundle gerbe connections are classified by degree 3 differential equivariant cohomology. The second theorem proves that a natural functor ' (1.3) G-Grbr(M) −! Grbr(ErG ×G M) is an equivalence of 2-groupoids. As explained in works such as [NS], a natural approach to equivariance in higher geometry is to use EG ×G M and the language of sheaves. The essential point of this paper is that when connections are involved, one should consider ErG ×G M instead of EG ×G M. As a sample application of Theorem 4.15, we consider the case where G is a compact semisimple Lie group acting on itself by conjugation. Theorem 5.4 gives a simple proof of the existence and uniqueness of certain G-equivariant gerbe connections on G. Such gerbes have been considered in several other works, including [BXZ, Bry2,GR, Mei,MW]. A CLASSIFICATION OF EQUIVARIANT GERBE CONNECTIONS 3 The paper is structured in the following way. Section2 contains a brief exposition of existing knowledge that will be utilized in the paper. The inclusion of this section serves the twofold purpose of establishing notations and conventions, and making the paper as self- contained as possible. Section3 contains an explicit definition of the 2-groupoid G-Grbr(M) and proves many basic properties, most of which follow easily from existing work in [Wal1, Mei, Sti]. Section4 introduces the second model Grbr(ErG ×G M), and we use the Weil homomorphism to define the functor (1.3). Making use of a result by Tu{Xu [TX], we are able to prove that (1.3) is a bijection on isomorphism classes, and we use the main result from [Red2] to prove that it induces equivalences of automorphism groupoids. Finally, Section5 contains a brief discussion of equivariant holonomy, and it proves an existence and uniqueness result for Ad-equivariant gerbes on compact semisimple Lie groups. We also explain a natural model for equivariant connections on p-gerbes for p > 1. Finally, it is worth noting that gerbes, both with and without connection, have played an important role in mathematical and physical problems, including twisted/differential K-theory and Ramond-Ramond field classifications [BCM+,KV, Par,GL], Wess-Zumino- Witten models [SSW], string structures [Wal2], and even topological insulators [Gaw,MT]. In each of these situations, it is natural to seek equivariant generalizations. The tools developed in this paper should aid especially in answering existence/uniqueness questions and in better developing functorial constructions. In particular, the formal properties of Grbr(ErG ×G M), displayed in (4.1), make equivariant gerbe connections excellent candi- dates for WZW terms in gauged sigma models [Wit, FOS, GSW1, GSW2,BE]. Acknowledgements. B.P. thanks Hausdorff Research Institute for Mathematics and Max Planck Institute for Mathematics for their support and hospitality during his visits. The authors thank Jim Stasheff for detailed comments on a previous draft. 2. Background and conventions We begin by recalling background information and explaining notation, most of which follows the conventions from [Red1, Red2,FH]. Throughout this paper, we work in the category Man of smooth manifolds with smooth maps. 2.1. Sheaves of 1-groupoids on the site of manifolds. A groupoid is a category whose morphisms are all invertible. Informally, an n-groupoid is an n-category where, for k ≥ 1, every k-morphism is invertible. Here, one works recursively and defines a k-morphism to be invertible if there exists some other k-morphism such that both compositions are isomorphic, via invertible (k + 1)-morphisms, to the respective identities. Though the structures we will study are naturally n-groupoids for n ≤ 2, it will be most convenient to let n = 1 and use the language of 1-groupoids, which are 1-categories whose k-morphisms are all invertible for k ≥ 1. Among the many excellent sources, see [Cam] for a useful introduction to 1-categories and [Lur] for complete details. The particular model we use for 1-Gpd, the 1-category of 1-groupoids, is the category of Kan complexes, viewed as a simplicially enriched full subcategory of simplicial sets. Most of our arguments, though, are model-independent and only require the informal notion of an 1-groupoid. Note that any set is naturally a groupoid with only the identity morphisms, and any groupoid determines a 2-groupoid with trivially defined 2-morphisms. We freely use these natural embeddings (fully faithful functors) Set ,! Gpd ,! 2-Gpd ,! 1-Gpd; 4 BYUNGDO PARK AND CORBETT REDDEN but we do not denote them by extra symbols. For example, the map Gpd ,! 1-Gpd is the nerve construction, but we usually avoid the symbol N and instead write Gpd ⊂ 1-Gpd. While this may be technically imprecise, we believe it is easier to read and will cause no confusion. Remark 2.1. Many of the categories we use, such as the category of principal G-bundles on a fixed manifold, are usually regarded as essentially small. The collections of objects do not necessarily form sets, but the categories are equivalent to small categories. In order to avoid technical problems, such as the nerve not forming a simplicial set, we assume that all objects are elements of some fixed Grothendieck universe, and therefore all categories in this paper are small. See [Lur, x1.2.15] for further discussion.